Chinese LLMs Broaden the Gap Between Attackers & Defenders

This article reports on the emergence of new Chinese AI models, GLM 5.2 and Tulongfeng (Dragon Saber), which are demonstrating strong performance in vulnerability discovery, rivaling leading US models like Opus and GPT-5.5. Experts warn that the increasing capabilities of these models, particularly their open-weight nature and cost-effectiveness, are significantly widening the gap between attackers and defenders, necessitating a rapid overhaul of security strategies to address vulnerabilities and prioritize patching. The rise of AI-powered vulnerability discovery is accelerating the threat landscape, demanding a shift in focus from model capabilities to robust defensive architectures and integration.